About This Project
Visual Jenga progressively removes objects from a scene to reveal implicit physical dependencies. Given a single image, the pipeline iteratively identifies and removes the "most removable" object using counterfactual inpainting until only the background remains.